Differentially Private Empirical Risk Minimization under the Fairness Lens

Neural Information Processing Systems 

It allows to measure and bound the risk associated with an individual participation in a computation.